The present invention relates to video-signal layered coding and decoding methods, apparatuses, and programs.
Several spatial- and temporal resolution and SNR (Signal-to-Noise) scalable video coding schemes have been proposed and employed in a variety of fields. In particular, spatial-domain resolution scalable video coding schemes are most applicable to still and moving pictures.
A spatial-resolution scalable layered video coding scheme produces interlayer predictive signals for use in coding enhancement layers that exhibit higher spatial resolutions than base layers also to be coded. The interlayer predictive signals are produced through interpolation of base-layer decoded signals.
In this coding scheme, an interlayer predictive signal is produced based on a correlation between an input enhancement-layer video signal and a base-layer signal. This interlayer correlation means that the base-layer signal carries some frequency components of the input enhancement-layer video signal, which further means that higher coding efficiency is achieved with higher correlation between the input enhancement-layer video signal and a base-layer decoded signal. Higher coding efficiency could be achieved with an interlayer predictive signal which is produced with an estimation procedure to give higher resolution to the base-layer decoded signal so that the input enhancement-layer video signal and the base-layer decoded signal have higher correlation, not produced with mere interpolation of the base-layer decoded signal.
When image enlargement is required, there is an image enlargement method for natural images with estimation of high-frequency components for higher resolution to an enlarged image, based on Laplacian pyramid in layered coding. Higher interlayer Laplacian-components correlation achieves prediction of Laplacian-components having spatial resolution of a coding layer from only a signal of a coding layer one layer lower than the former layer.
The above image enlargement method causes several problems if it is directly applied to the estimation procedure in layered coding discussed above, because it is for natural images. A base-layer decoded signal is a coarse signal with few components of high-frequency components that are carried by an input enhancement-layer video signal. Moreover, coarse quantization gives lower correlation between the base-layer decoded signal and the input enhancement-layer video signal. The above image enlargement method thus does not assure desired high coding efficiency if it is directly applied to the estimation procedure in layered coding, discussed above, with estimation of high-frequency components from a given low-resolution signal only.